Creating an Interactive Correlation Plot with Filtering Capabilities

Research Question

What are we analyzing?
We aim to create an interactive correlation plot using plotly that allows filtering to observe how correlations change across different groups (e.g., species in the Iris dataset).

Step 1: Loading Libraries and Data

Loads the required libraries and the Iris dataset for analysis.

library(plotly)
library(dplyr)

data(iris)
head(iris)
##   Sepal.Length Sepal.Width Petal.Length Petal.Width Species
## 1          5.1         3.5          1.4         0.2  setosa
## 2          4.9         3.0          1.4         0.2  setosa
## 3          4.7         3.2          1.3         0.2  setosa
## 4          4.6         3.1          1.5         0.2  setosa
## 5          5.0         3.6          1.4         0.2  setosa
## 6          5.4         3.9          1.7         0.4  setosa

Step 2: Defining Regression and Correlation Line Function

Creates a custom function to calculate regression lines and correlation coefficients for subsets of data (based on species

regression_lines <- function(data, species) {
  data_sp <- data %>% filter(Species == species)
  model <- lm(Petal.Length ~ Sepal.Length, data = data_sp) 
  
  x_range <- seq(min(data_sp$Sepal.Length), max(data_sp$Sepal.Length), length.out = 100)
  y_range <- predict(model, newdata = data.frame(Sepal.Length = x_range))
  
  cor_value <- round(cor(data_sp$Sepal.Length,
                         data_sp$Petal.Length), 2) # Calculate correlation coefficient
  
  data.frame(Sepal.Length = x_range, Petal.Length = y_range, Species = species, Correlation = cor_value)
}

Step 3: Generating Regression Lines and Correlation Coefficients

Generates regression lines and calculates correlation coefficients for each species (Setosa, Versicolor, Virginica).

lines_setosa <- regression_lines(iris, "setosa")
lines_versicolor <- regression_lines(iris, "versicolor")
lines_virginica <- regression_lines(iris, "virginica")

Step 4: Creating the Interactive Plot with Data Points

Creates a scatter plot of Sepal.Length vs Petal.Length and includes a filter for species.

fig <- plot_ly(
  data = iris,                                              # Data source
  x = ~Sepal.Length,                                       # X-axis: Sepal Length
  y = ~Petal.Length,                                       # Y-axis: Petal Length
  color = ~Species,                                        # Color by Species
  colors = c('blue', 'orange', 'green'),                   # Define colors for each species
  type = 'scatter',                                        # Specify the plot type as scatter
  mode = 'markers',                                        # Display markers
  transforms = list(
    list(
      type = 'filter',                                      # Add a filter transform
      target = ~Species,                                    # Target the Species variable
      operation = '=',                                      # Operation type: equals
      value = "setosa"                                      # Initial filter value: setosa
    )
  )
)

Step 5: Adding Regression Lines to the Plot

Adds dynamic regression lines for each species to the scatter plot.

fig <- fig %>%
  add_lines(
    data = lines_setosa,                                      # Data for Setosa regression line
    x = ~Sepal.Length,                                       # X values for the line
    y = ~Petal.Length,                                       # Y values for the line
    line = list(color = 'blue', width = 1),                  # Line style
    name = paste("Setosa (r =", lines_setosa$Correlation[1], ")")  # Legend name with correlation
  ) %>%
  add_lines(
    data = lines_versicolor,                                  # Data for Versicolor regression line
    x = ~Sepal.Length,                                       # X values for the line
    y = ~Petal.Length,                                       # Y values for the line
    line = list(color = 'orange', width = 1),                # Line style
    name = paste("Versicolor (r =", lines_versicolor$Correlation[1], ")")  # Legend name with correlation
  ) %>%
  add_lines(
    data = lines_virginica,                                   # Data for Virginica regression line
    x = ~Sepal.Length,                                       # X values for the line
    y = ~Petal.Length,                                       # Y values for the line
    line = list(color = 'green', width = 1),                 # Line style
    name = paste("Virginica (r =", lines_virginica$Correlation[1], ")")  # Legend name with correlation
  )

Step 6: Adding Filtering Capabilities to the Plot

Includes a dropdown menu to filter the plot by species or show all data points together.

fig <- fig %>%
  layout(
    title = "Dynamic Lines with Correlation Coefficients",          # Plot title
    xaxis = list(title = "Sepal Length (cm)"),                      # X-axis title
    yaxis = list(title = "Petal Length (cm)"),                      # Y-axis title
    updatemenus = list(
      list(
        buttons = list(
          list(
            method = "restyle",                                     # Method to update the plot
            args = list("transforms[0].value", "setosa"),          # Filter for Setosa
            label = "Iris-setosa"                                   # Button label
          ),
          list(
            method = "restyle",                                     # Method to update the plot
            args = list("transforms[0].value", "versicolor"),      # Filter for Versicolor
            label = "Iris-versicolor"                               # Button label
          ),
          list(
            method = "restyle",                                     # Method to update the plot
            args = list("transforms[0].value", "virginica"),       # Filter for Virginica
            label = "Iris-virginica"                                # Button label
          ),
          list(
            method = "restyle",                                     # Method to update the plot
            args = list("transforms[0].value", unique(iris$Species)),  # Show all species
            label = "All"                                           # Button label
          )
        ),
        direction = "down",                                           # Dropdown direction
        x = 0.1,                                                      # X position of the dropdown
        y = 1.15,                                                     # Y position of the dropdown
        showactive = TRUE                                             # Show active button
      )
    )
  )

fig

About the Plot

What the plot shows:
This interactive plot visualizes the relationship between Sepal.Length and Petal.Length for the Iris dataset, with regression lines and correlation coefficients for each species. Key features include:

  1. Filtering:
    • Use the dropdown menu to filter the data by species (Setosa, Versicolor, or Virginica) or view all species together.
  2. Dynamic Regression Lines:
    • Each species has its regression line displayed with its correlation coefficient (r). These lines dynamically adjust based on the selected filter.
  3. Interactive Legend:
    • You can click on the legend items to toggle the visibility of specific data points or regression lines, allowing for customized views.
  4. Scalability:
    • The plot supports zooming and panning, making it easy to focus on specific data points or regions of interest.
    • Use the built-in toolbar to reset the view, save the graph, or activate zoom modes.

This plot is an excellent tool for exploring both individual and combined group correlations, enabling deeper insights into the data structure.

Conclusion

Key Insights:

  1. Interactive Filtering and Group Analysis:
    • The dropdown menu allows users to isolate specific species or compare all groups together, providing flexibility in exploring correlations.
  2. Dynamic Regression Lines and Correlations:
    • The regression lines, annotated with r values, provide a clear understanding of the strength and direction of linear relationships for each group.
    • This makes it easy to compare correlations across different subsets.
  3. Legend Interactions and Zooming:
    • The interactive legend enables users to toggle data visibility, simplifying the exploration of specific subsets or regression lines.
    • Zooming and panning features improve usability, especially when analyzing dense or overlapping data points.

Recommendation:
This approach is ideal for datasets with well-defined groups (e.g., categories or classes). Use dynamic filtering to explore correlations efficiently across subsets. The interactive legend and zoom features further enhance the user experience, making it suitable for exploratory data analysis and presentations to diverse audiences.